TASK ABORTED status clarification - 06-026r2.pdf

Charles Binford Charles.Binford at Sun.COM
Mon Jun 25 09:51:28 PDT 2007


Formatted message: <A HREF="r0706252_f.htm">HTML-formatted message</A>

Rob, my current position within Sun no longer has me dealing with T10
issues, but since the Task Aborted status came from me way back when, I
thought I'd give a couple of comments on the proposal via the reflector.
>From your proposal
Overview
	It is not clear whether TASK ABORTED status is sent when tasks
	are aborted by a LOGICAL UNIT RESET
	task management function.
I could argue that SAM-3 and SAM-4 clearly indicate that the TASK
ABORTED status is to be used when tasks on another I_T_nexus are aborted
do to a LOGICAL UNIT RESET.  But since you are already going in the
"right" direction with your proposal that argument would be a waste of
time.  Instead, let me just re-emphasize (in case there is discussion to
the contrary during CAP) that as the original author of the Task Aborted
concept, it was my intent from the beginning that this status would
apply to LOGICAL UNIT RESET.  In fact the notification issue after, LUR
and CLEAR TASK SET were the driving force behind the proposal in the
first place.
Now on to specific comments about the proposed changes.
Table 27, Note 'c' for the Hard Reset row
I don't understand note 'c' and the 'yes or no' text in the the table.
Looking at the referenced text in SAM (6.3.2) I see a couple of relevant
phrases:
	the logical unit reset condition established by a hard reset may
	affect tasks that are communicating via other SCSI target ports.
and
	A SCSI port’s response to a hard reset condition shall include
	establishing an I_T nexus loss condition (see 6.3.4) for every
	I_T nexus associated with that SCSI port.
I think this table is trying to say that TASK ABORTED should be used by
tasks on other SCSI target ports that get aborted because of the logical
unit reset established by the hard reset, but tasks that are aborted on
a port that is going to report an I_T nexus loss do not get a TASK
ABORTED status.  This makes sense to me but the table/note don't say it
very clearly yet.
Tables 28 and 29 Unit Attention and Task Aborted columns
The "COMMANDS CLEARED BY ANOTHER INITIATOR" unit attention and the TASK
ABORTED STATUS are mutually exclusive based on the setting of the TAS
bit.  This fact is not clear in the table.  The note 'a' in both tables
for Task Abort status has the 'if enable' tacked on but there is no
indication on the Unit Attention.  A similar note for the Unit Attention
column (or at least on the rows with "COMMANDS CLEARED BY ANOTHER
INITIATOR" would help, but I'd think it would be better if mutual
exclusive nature of the "method of notification" (to use the term used
in SAM-4 5.6.3) was more prominent in the table.
Thanks,
Charles Binford



More information about the T10 mailing list